Disclosed are systems and methods to identify text-like pixels from an image by providing an image and classifying line segments of pixels within the image by edge-bounded averaging.
Legal claims defining the scope of protection, as filed with the USPTO.
1. A method to identify text-like pixels from an image, the method comprising: classifying a plurality of individual pixels within a mask within the image as either edge or non-edge, wherein a pixel (i,j) is located at the center of the mask; determining whether the pixel (i,j) is an edge pixel or a non-edge pixel; determining pixels in the mask having connectivity with the pixel (i,j); for each pixel in the mask determined to have connectivity with the pixel (i,j), determining whether the pixel is an edge pixel or a non-edge pixel; and performing, by a computer, edge-bounded averaging to determine line segments, wherein the edge-bounded averaging comprises one of in response to determining that pixel (i,j) is an edge pixel, identifying all pixels in the mask that are edge pixels and that have connectivity with the pixel (i,j) from the determining of whether each pixel having connectivity is an edge pixel or a non-edge pixel, and determining an average value of only all the identified edge pixels in the mask having connectivity with the pixel (i,j), or in response to determining that pixel (i,j) is a non-edge pixel, identifying all pixels in the mask that are non-edge pixels and that have connectivity with the pixel (i,j) from the determining of whether each pixel having connectivity is an edge pixel or a non-edge pixel, and determining an average value of only all the identified non-edge pixels in the mask having connectivity with the pixel (i,j).
2. The method of claim 1 , further comprising: examining sub-blobs of pixels within the image; and performing sub-blob connectivity analysis.
3. The method of claim 2 , further comprising: identifying and classifying edges of pixels within the image; performing filling to further classify pixels within the image; performing consistency analysis of pixels within the image; performing pixel connectivity analysis of pixels within the image; and identifying text pixels within the image.
4. The method of claim 2 , wherein the step of examining sub-blobs of pixels within the image comprises: examining each sub-blob to determine whether it is NON TEXT.
5. The method of claim 1 , further comprising performing color space conversion of the image.
6. The method of claim 1 , further comprising smoothing the image.
7. The method of claim 1 , wherein a Gaussian lowpass filter is applied to the image, the filter being f i , j = k ⅇ - α 2 [ ( ⅈ - c ) 2 + ( j - c ) 2 ] / c 2 where k is a normalizing factor such that ∑ i , j f i , j = 1.0 , c is the center of the filter and β=1.0.
8. The method of claim 3 , wherein the step of identifying and classifying edges of pixels within the image further comprises, classifying every pixel as NON EDGE, WHITE EDGE or BLACK EDGE.
10. The method of claim 1 , wherein the step of performing edge-bounded averaging further comprises: starting from a first side of a line proceeding to a second side of the line identifying consecutive segments of pixels as NON EDGE, WHITE EDGE or BLACK EDGE.
11. The method of claim 1 , wherein the step of performing edge-bounded averaging comprises: computing the edge-bounded averaging for at least eight locations including both end points of a central interior, both end points of a left edge segment, both end points of a right edge segment, a right end point of a left interior and a left end point of a right interior.
12. The method of claim 11 , further comprising: classifying the central interior as NON TEXT, BLACK INTERIOR or WHITE INTERIOR based upon the edge-bounded averaging values.
13. The method of claim 3 , wherein the step of performing filling to further classify pixels within the image comprises: classifying segments as NON TEXT; and examining segments classified as NON TEXT to determine whether they may be reclassified as BLACK INTERIOR, BLACK EDGE, WHITE INTERIOR or WHITE EDGE.
14. The method of claim 3 , wherein the step of performing vertical consistency analysis of pixels within the image comprises: examining pixels not yet classified as NON TEXT to determine whether they are BLACK INTERIOR, BLACK EDGE, WHITE INTERIOR or WHITE EDGE.
15. The method of claim 3 , wherein the step of performing pixel connectivity analysis of pixels within the image comprises: identifying aggregates of pixels having been identified as candidates for text, the aggregates being sub-blobs; and collecting statistics with respect to each sub-blob, wherein said statistics are selected from the group consisting of total number of pixels, sums of color values, number of border pixels, number of broken border pixels and horizontal run length.
16. The method of claim 3 , wherein the step of identifying text pixels comprises: examining each sub-blob to classify each pixel as either a text pixel or a non-text pixel.
17. A system for identifying text-like pixels from an image, the system comprising: a processor for classifying a plurality of individual pixels within a mask within the image as either edge or non-edge, wherein a pixel (i,j) is located at the center of the mask; determining whether the pixel (i,j) is an edge pixel or a non-edge pixel; determining pixels in the mask having connectivity with the pixel (i,j); for each pixel in the mask determined to have connectivity with the pixel (i,j), determining whether the pixel is an edge pixel or a non-edge pixels and performing edge-bounded averaging to determine line segments, wherein the edge-bounded averaging comprises one of in response to determining that pixel (i,j) is an edge pixel, identifying all pixels in the mask that are edge pixels and that have connectivity with the pixel (i,j) from the determining of whether each pixel having connectivity is an edge pixel or a non-edge pixel, and determining an average value of only all the identified edge pixels in the mask having connectivity with the pixel (i,j), or in response to determining that pixel (i,j) is a non-edge identifying all pixels in the mask that are non-edge pixels and that have connectivity with the pixel (i,j) from the determining of whether each pixel having connectivity is an edge pixel or a non-edge pixel, and determining an average value of only all the identified non-edge pixels in the mask having connectivity with the pixel (i,j).
18. The system of claim 17 , wherein the processor also examines sub-blobs of pixels within the image; and performs sub-blob connectivity analysis.
19. The system of claim 18 , wherein the processor also identifies and classifies edges of pixels within the image; performs vertical filling to further classify pixels within the image; performs vertical consistency analysis of pixels within the image; performs pixel connectivity analysis of pixels within the image; and identifies text pixels.
20. A non-transitory computer readable storage medium on which is embedded one or more computer programs comprising a set of instructions that when executed by a processing circuit performs a method of processing a digital image, the method comprising: classifying a plurality of individual pixels within a mask within the digital image as either edge or non-edge, wherein a pixel (i,j) is located at the center of the mask; determining whether the pixel (i,j) is an edge pixel or a non-edge pixel; determining pixels in the mask having connectivity with the pixel (i,j); for each pixel in the mask determined to have connectivity with the pixel (i,j), determining whether the pixel is an edge pixel or a non-edge pixel; and performing, by a computer, edge-bounded averaging to determine line segments, wherein the edge-bounded averaging comprises one of in response to determining that pixel (i,j) is an edge pixel, identifying all pixels in the mask that are edge pixels and that have connectivity with the pixel (i,j) from the determining of whether each pixel having connectivity is an edge pixel or a non-edge pixel, and determining an average value of only all the identified edge pixels in the mask having connectivity with the pixel (i,j), or in response to determining that pixel (i,j) is a non-edge pixel, identifying all pixels in the mask that are non-edge pixels and that have connectivity with the pixel (i,j) from the determining of whether each pixel having connectivity is an edge pixel or a non-edge pixel, and determining an average value of only all the identified non-edge pixels in the mask having connectivity with the pixel (i,j).
21. The non-transitory computer readable storage medium according to claim 20 , said one or more computer programs further comprising a set of instructions for: performing pixel connectivity analysis of pixels within the digital image identifying aggregates of pixels having been identified as candidates for text, the aggregates being sub-blobs; collecting each sub-blobs statistics: total number of pixels, sums of color values, number of border pixels, number of broken border pixels and horizontal run length; and counting sums of each luminance and chroma.
22. The non-transitory computer readable storage medium according to claim 20 , said one or more computer programs further comprising a set of instructions for: performing pixel connectivity analysis of pixels within the digital image by identifying aggregates of pixels having been identified as candidates for text, the aggregates being sub-blobs; collecting each sub-blobs statistics: total number of pixels, sums of color values, number of border pixels, number of broken border pixels and horizontal run length; and counting sums of each Y, C r , C b .
Cooperative Patent Classification codes for this invention. Click any code to explore related patents in that topic.
January 11, 2002
January 24, 2012
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.